对py自升级
利用海康vm的文本行识别
得出文本行的rol矩形框，
导入python脚本，
python脚本塞入参考
C:\Program Files\VisionMaster4.4.0\Applications\ModuleProxy\x64
python
import sys
sys.argv=['']
from ioHelper import *
import cv2 as cv
import numpy as np
def Process(data) -> int:
moduleVar = IoHelper(data, INIT_MODULE_VAR)
globalVar = IoHelper(data, INIT_GLOBAL_VAR)
localVar = IoHelper(data, INIT_LOCAL_VAR)
try:
# 只保留灰度图输入（彩色全部删除）
imagedata_gray = moduleVar.in3 # 灰度图像源
width = moduleVar.in1 # 图像宽
height = moduleVar.in2 # 图像高
# 灰度图字节流转 numpy
img_np_gray = np.frombuffer(imagedata_gray.buffer, np.uint8)
gray = np.reshape(img_np_gray, [height, width]) # 直接重塑为灰度图
# 二值化
ret, img_threshold = cv.threshold(gray, 127, 255, cv.THRESH_BINARY)
# 轮廓检测
contours, hierarchy = cv.findContours(img_threshold, cv.RETR_TREE, cv.CHAIN_APPROX_SIMPLE)
# 在灰度图上绘制红色轮廓（保持视觉效果）
img_contour = cv.cvtColor(gray, cv.COLOR_GRAY2BGR) # 转彩色用于画轮廓
cv.drawContours(img_contour, contours, -1, (0, 0, 255), 2)
img_contour_gray = cv.cvtColor(img_contour, cv.COLOR_BGR2GRAY) # 转回灰度
# 腐蚀（图像收缩）
kernel = np.ones((5, 5), np.uint8)
erosion = cv.erode(img_threshold, kernel, iterations=1)
# 输出：腐蚀后的灰度图
img_out1 = imagedata_gray
img_out1.buffer = erosion.tobytes()
moduleVar.out0 = img_out1
# 输出：带轮廓的灰度图
img_out2 = imagedata_gray
img_out2.buffer = img_contour_gray.tobytes()
moduleVar.out1 = img_out2
except BaseException as e:
PrintMsg(e)
return 0python -m pip install --upgrade pip setuptools wheel -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com
为嵌入的python添加指定的库
pip install opencv-python
pip 安装：pip install opencv-python（基础功能）
或pip install opencv-contrib-python（包含额外模块）
安装时包名为opencv-python，但导入时使用import cv2
常用操作与代码示例
图像读取与显示：
读取：cv2.imread('image.jpg')，默认采用 BGR 通道顺序。27
显示：cv2.imshow('window', img)配合cv2.waitKey(0)使用。7
保存：cv2.imwrite('output.jpg', img)3
图像处理基础：
颜色转换：cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)将彩色图转为灰度图。25
图像缩放：cv2.resize(img, (width, height))3
图像翻转：cv2.flip(img, flipcode)，flipcode=0 垂直翻转，1 水平翻转，-1 同时翻转。3
滤波与边缘检测：
高斯滤波：cv2.GaussianBlur(img, (5,5), 0)去除高斯噪声。37
中值滤波：cv2.medianBlur(img, 5)适合处理椒盐噪声。3
Canny 边缘检测：cv2.Canny(gray, 100, 200)提取图像边缘。37转成除矩形框外全部白色
python
import cv2
import numpy as np
def keep_multi_roi_white_bg(image_path, output_path, roi_list):
"""
多个ROI区域保留原图，其余全部变白
:param image_path: 输入图片路径
:param output_path: 输出保存路径
:param roi_list: 列表嵌套，每个元素是一个四边形四点坐标
格式: [[[x1,y1],[x2,y2],[x3,y3],[x4,y4]], 第二个ROI, ...]
"""
img = cv2.imread(image_path)
h, w = img.shape[:2]
# 初始化全白画布
white_bg = np.ones_like(img) * 255
# 初始化掩码
mask = np.zeros((h, w), dtype=np.uint8)
# 遍历所有ROI，逐个填充掩码
for roi_pts in roi_list:
pts = np.array(roi_pts, dtype=np.int32)
cv2.fillPoly(mask, [pts], 255)
# 掩码区域贴原图，其余保持白色
white_bg[mask == 255] = img[mask == 255]
cv2.imwrite(output_path, white_bg)
print("多ROI处理完成，已保存！")
if __name__ == "__main__":
# ========== 只需要修改这里 ==========
INPUT = "input.png"
OUTPUT = "output.png"
# 多个ROI示例：两个长方形
ROI_LIST = [
# 第一个ROI四点
[
[50, 50],
[300, 50],
[300, 200],
[50, 200]
],
# 第二个ROI四点
[
[350, 80],
[600, 80],
[600, 250],
[350, 250]
]
# 你可以继续添加第三个、第四个...
]
# =================================
keep_multi_roi_white_bg(INPUT, OUTPUT, ROI_LIST)
python
import sys
sys.argv=['']
from ioHelper import *
import cv2 as cv
import numpy as np
def Process(data) -> int:
moduleVar = IoHelper(data, INIT_MODULE_VAR)
globalVar = IoHelper(data, INIT_GLOBAL_VAR)
localVar = IoHelper(data, INIT_LOCAL_VAR)
try:
# 获取灰度图输入
imagedata_gray = moduleVar.in3 # 黑白图像
width = moduleVar.in1 # 宽
height = moduleVar.in2 # 高
# 字节流转灰度图
img_np_gray = np.frombuffer(imagedata_gray.buffer, np.uint8)
gray = np.reshape(img_np_gray, [height, width])
# ===================== 你只需要改这里的ROI坐标 =====================
ROI_LIST = [
[[50, 50], [300, 50], [300, 200], [50, 200]], # 第一个区域
[[350, 80], [600, 80], [600, 250], [350, 250]], # 第二个区域
# 可继续加第三个、第四个...
]
# 创建全白背景
white_bg = np.ones_like(gray) * 255
mask = np.zeros((height, width), dtype=np.uint8)
# 填充所有ROI区域
for roi_pts in ROI_LIST:
pts = np.array(roi_pts, dtype=np.int32)
cv.fillPoly(mask, [pts], 255)
# ROI保留原图，其余变白
white_bg[mask == 255] = gray[mask == 255]
# 输出结果（out0 和 out1 输出同一张图）
img_out = imagedata_gray
img_out.buffer = white_bg.tobytes()
moduleVar.out0 = img_out
moduleVar.out1 = img_out
except BaseException as e:
PrintMsg(e)
return 0在其他ocr里进行识别（嵌入式的py受限）
ocr识别
ocr识别 tessera
百度飞桨
python
import sys
sys.argv=['']
from ioHelper import *
import cv2 as cv
import numpy as np
def Process(data) -> int:
moduleVar = IoHelper(data, INIT_MODULE_VAR)
globalVar = IoHelper(data, INIT_GLOBAL_VAR)
localVar = IoHelper(data, INIT_LOCAL_VAR)
try:
imagedata_gray = moduleVar.in0
width = moduleVar.in1
height = moduleVar.in2
img_np_gray = np.frombuffer(imagedata_gray.buffer, np.uint8)
gray = np.reshape(img_np_gray, [height, width])
# ===================== 从变量读取多ROI Box =====================
center_x = moduleVar.in3.center_x
center_y = moduleVar.in3.center_y
box_w = moduleVar.in3.width
box_h = moduleVar.in3.height
white_bg = np.ones_like(gray) * 255
mask = np.zeros((height, width), dtype=np.uint8)
# 批量生成ROI
for cx, cy, w, h in zip(center_x, center_y, box_w, box_h):
x1 = int(cx - w / 2)
y1 = int(cy - h / 2)
x2 = int(cx + w / 2)
y2 = int(cy + h / 2)
roi_pts = [[x1, y1], [x2, y1], [x2, y2], [x1, y2]]
pts = np.array(roi_pts, dtype=np.int32)
cv.fillPoly(mask, [pts], 255)
white_bg[mask == 255] = gray[mask == 255]
img_out = imagedata_gray
img_out.buffer = white_bg.tobytes()
moduleVar.out0 = img_out
moduleVar.out1 = img_out
except BaseException as e:
PrintMsg(e)
return 0